Editorial policies

How papers are assessed, how conflicts are handled, and what happens when something published here turns out to be wrong.

Computational Manuscript Review (CMR)

Every manuscript submitted to the journal undergoes a Computational Manuscript Review (CMR): a structured review conducted by the most advanced frontier AI Large Language Models under the journal's review protocol. The review evaluates originality, methodological rigor, evidentiary support, citation accuracy and clarity, and puts four questions to every paper: does its method support its claims; are its numbers reproducible from the data as presented; are its limitations stated honestly; and does it claim more than it has shown.

Each review is then examined by at least one human editor, who reads it alongside the manuscript and decides whether the submission meets the journal's standards for publication. Authors may be required to address the deficiencies the review identifies before acceptance. Accepted articles carry the designation CMR · Editorially Accepted: Computational Manuscript Review, Human Editorial Approval. The process is distinct from conventional external peer review, and final responsibility for acceptance and publication rests with the journal's editors. Publication is not a statement that the editors agree with the conclusions.

Why the journal uses computational review

Conventional peer review remains the default standard of scholarly publishing, but its limitations are well documented, among them low agreement between reviewers, bias associated with authors, topics and methods, and resistance to findings that depart from prevailing views. The most direct current evidence on an alternative comes from a 2026 study at Harvard University's Institute for Quantitative Social Science, which compared AI-generated reviews with reviews written for the same manuscripts by well-compensated expert reviewers, reviews that the authors rated highly and that former journal editors described as among the best they had seen. In blind assessments by an independent AI judge and by former editors of major journals, the AI reviews were judged better on every criterion the researchers measured; they cited no fabricated sources and made incorrect claims about the manuscripts roughly a third as often as the human reviews (King & Worthington, 2026). The same study concluded that publication decisions should remain with human editors, which is the division this journal adopts.

These considerations weigh especially heavily in American kava research. The field is small and recent; many of the people qualified to review its work hold commercial, institutional or professional positions in the questions it studies; and a number of its findings depart from assumptions established in the earlier literature. In such a field, review can easily become gatekeeping rather than assessment. Computational review applies the same criteria to every manuscript and has no stake in the conclusions it examines. It has limitations of its own, since language models can reflect the conventions of the literature on which they were trained, and for that reason every review is read by a human editor, who makes the decision.

Reference: King, G., & Worthington, S. (2026). AI article reviews dominate human peer reviews at their best: Evidence and policy choices [Working paper]. Institute for Quantitative Social Science, Harvard University. https://garyking.org/peerless

Publication criteria

Papers are not rejected for being negative, inconclusive, small, or for reporting that an earlier claim — including one published in this journal — was wrong. Underpowered work is publishable if it says it is underpowered, and is labeled a working paper.

Editors as authors

Editors are authors on a substantial share of what the journal publishes, and that is disclosed on every such paper rather than managed out of sight. Their papers go through Computational Manuscript Review like any other. Where an editor is an author, the decision is taken by another editor wherever possible, and the article page names the handling editor; where the handling editor is also an author, the article page says so, so that a reader knows exactly what stands behind it.

Conflicts of interest

Every paper carries a conflicts-of-interest statement, and "none" is a statement that must be earned. Authors must disclose ownership or employment in any business that grows, imports, processes, tests or sells kava or kava products; ownership of relevant plant material, cultivars, trademarks, patents or plant variety rights; any financial relationship with a party whose product is evaluated; and any role in a standards or certification body whose framework the paper applies or endorses.

Disclosure is not disqualifying. Concealment is: a paper found to have omitted a material interest is corrected with an editorial note, and if the omission is not correctable, retracted.

Authorship

Authorship requires substantial contribution to the design, conduct, analysis or writing of the work, plus approval of the final text and willingness to be accountable for it. Supplying material, funding the work, or running a commissioned analysis does not by itself earn authorship; those contributions belong in the acknowledgments. Contract laboratories are named in the methods with their report or batch identifier, not listed as authors.

Changes to the author list after submission require the written agreement of everyone involved, including anyone being removed.

Data and materials

Papers must state where their supporting data are and on what terms they can be obtained. The journal's default — consistent with the American Kava Association's position that its self-funded research serves the growers who paid for it — is that primary data are held by the corresponding author and made available to stakeholders in the American kava growing industry on request. A paper may adopt a more open position, and several do.

Whatever the access terms, a paper must present enough of its own data, analytical provenance and replicate structure for a reader to check its arithmetic. Conclusions resting on numbers that appear nowhere in the paper are not publishable. The For authors page sets out what that means in practice.

Versions, corrections and retraction

Articles are not silently edited. Typographical fixes that do not touch meaning are made without notice. Anything else produces a dated note on the article page and, where the change affects a finding and once DOI deposits begin, a formal update deposited with Crossref so that anyone holding the DOI learns of it.

1. Version numbering. Exploratory releases carry a letter suffix (v1.0a) and are labeled Working Paper on their article page. They are citable, and they are expected to be superseded by later data. A working paper that has been superseded links forward to the version that replaced it.
2. Correction. An error that does not invalidate the conclusions. The corrected text stands, with a dated correction note at the head of the article stating exactly what changed.
3. Retraction. An error, omission or misconduct that invalidates the conclusions. The article stays online with a retraction notice across it; it is never deleted, because deleting it would break every citation to it.
4. Withdrawal of a claim. Where authors themselves withdraw a specific claim on later evidence, the withdrawal is reported near the top of the paper rather than buried, and is treated as a normal scientific act rather than an embarrassment.

Declaration of AI use

Authors must declare any use of generative AI in preparing a manuscript, saying what it was used for — drafting, editing, code, figure preparation, literature search — and confirming that the authors verified the output. AI systems are not authors and cannot be listed as such. Undeclared use discovered after publication is handled as a correction. In assessing manuscripts, the journal's own use of AI is confined to Computational Manuscript Review, described above; editorial decisions are made by human editors.

Originality and duplicate publication

Submissions must be original and not under consideration elsewhere. Posting a preprint or circulating a technical white paper before submission is not duplicate publication; the journal will consider work that has previously appeared as an association white paper, provided the earlier version is cited on the article page and the relationship between the two is stated.

An article published here is the article of record. If you plan to publish related work in another journal, tell the editors before acceptance.

Republished guest papers are marked as such, reproduced unedited with the author's written permission, carry the author's own affiliation rather than the journal's, and separate any editorial commentary from the paper itself.

Complaints and appeals

Anyone may write to the editor to dispute a published finding or the handling of a paper. Substantive disputes about a finding are, where possible, handled in public: the journal would rather publish a well-argued comment and the authors' reply than settle an argument by correspondence.

Ethics involving people and communities

Work involving human participants — sensory panels, consumption studies, surveys, interviews — must report how consent was obtained and what the participants were told. Ethnographic and historical work concerning Pacific Island communities must represent those communities' own accounts accurately and cite Pacific scholarship where it exists, including scholarship critical of commercial kava outside the Pacific. Disagreeing with a critique is acceptable; failing to engage with it is not.